Energy Deviation Values of Gabor Filters in Texture Classification

J. Dourado, J. Rouco, M.G. Penedo, M. Ortega, and A. Mosquera (Spain)


Texture description, Texture classification, Gabor filtering.


Texture is one of the most important visual characteristics for image analysis. Several texture feature extraction methods proposed in the literature agree in the use of statistical metrics of a filter bank output along local spatial regions. Among these, the most commonly used descriptors are those based in the combination of averages and standard deviations of the magnitude responses of the filter bank. In this paper, we are concerned with the quantitative evaluation of the performance improvement associated to the combined use of average and standard deviation of energy values in comparison with the only use of energy average values. Our results show that the performance improvement although not very large in average, is actually very significant for some kinds of textures.

Important Links:

Go Back